Computational Advertising Recommendation System Based on Graph Neural Network

Computational advertising is an advertising format that determines the most matching advertisement according to specific users and specific situations through efficient algorithms and carries out precise creation, production, delivery, dissemination and interaction. As an effective way to computational advertising, the recommendation system can realize the accurate delivery of customer groups. The existing recommendation system runs by predicting the scores of unrated items and analyzing large-scale data sets. For accurate recommendation, the relevant information of each user and item needs to be considered. In this process, a lot of computing resources are required and the operation complexity is high. In the process of building the recommendation system, if only the influence diffusion process in the social network is modeled, the potential cooperative interest of users hidden in the user-item interest network will be ignored. Therefore, this paper combines the purchase record information and rating information of users' personal information with the intention information of horizontal similar users, and proposes a recommendation system model based on graph neural network, which takes into account user behavior and social relationship data and user personality information, optimizes recommendation results, and realizes efficient recommendation process and personalized recommendation results. The three public data sets of Ciao, Epinions and UserBehavior were compared with the optimal baseline model, and the average absolute error (MAE), root mean square error (RMSE) Recall and Precision were used as the evaluation indexes of the model. The experimental results show that the proposed method performs better than other methods in the model based on user relationship.